Interrupted out
When the host cuts a crew run off five times (its lease lapses), each time the task is queued again after a wait of 15 s, doubling to 4 min; the count is the run's interruptions column. The sixth time the run is marked failed with "the host stopped this run five times; not tried again". A graceful stop for a restart neither counts nor waits. Runs failed under the older "stopped twice by its forge restarting" text still show. Home names each such run (last 24 hours, up to five, repositories you may write) in one line with Try again; so does the terminal.
- Web: Home, under the held line. Try again posts (same origin, signed in) to
/:owner/:name/runs/:id/requeue, which queues the run's task again and returns to Home. Writers with the crew ability only; others are refused. - Terminal: the palette's "Stopped runs".
j/kmove,Enterorrtries again. The screen reads the database on every draw, so a retry on either door shows on the other. - Gone when: a newer agent run of the same issue exists, the run is over a day old, or you cannot write the repository.
- Both doors read
interruptedOutRuns(src/runs/interrupted_out.ts) and write throughrequeueRepoRun. The forge may refuse a retry (one already waiting, the day's ceiling); it says so.